Alluvial riverbed elevation responds to the balance between sediment supply and transport capacity, which is largely dependent on climate and its translation into fluvial discharge. We examine these relations using U.S. Geological Survey streamflow and channel measurements in conjunction with basin characteristics for 915 reference (“least disturbed”) measurement stations across the conterminous United States for the period A.D. 1950–2011. We find that (1) 68% of stations have bed elevation change (BEC) trends (p < 0.05) with median values of +0.5 cm/yr for aggradation and –0.6 cm/yr for degradation, with no obvious relation to drainage basin structure, physiography, or lithology; (2) BEC correlates with drainage basin area; (3) high-flow variability (Q90/Q50, where Q is discharge and 90 and 50 are annual flow percentiles) translates directly into the magnitude, though not the direction, of BEC, after accounting for the scale dependence; (4) Q90/Q50 declines systematically from dry to wet climates, producing disproportionately high rates of BEC in drier regions; and (5) marked increases in precipitation and streamflow occurred disproportionately at dry sites, while streamflow declined disproportionately at wet sites. Climatic shifts in streamflow have the potential to increase/decrease sediment flux and thus affect riverbed elevation by altering flood frequency. These unforeseen responses of bed elevation to climate and climate change have important implications for sediment budgets, longitudinal profiles, ecology, and river management.

Riverbed elevation is a fundamental boundary condition for landscape evolution (Tucker and Slingerland, 1997), alluvial stratigraphy (Daniels, 2008), flood risk, and aquatic habitat (ASCE Task Committee on Sediment Transport and Aquatic Habitats, Sedimentation Committee, 1992). Over societally relevant time scales, anthropogenic activity (Syvitski et al., 2005) and climate (Inman and Jenkins, 1999) govern changes in sediment production and flux through fluvial systems, and by extension, they affect alluvial riverbeds (Daniels, 2008; Simon, 1989; Williams and Wolman, 1984). But in contrast to substantial work documenting anthropogenic impacts in alluvial rivers (Gregory, 2006), we have no corresponding understanding of climatic riverbed signatures, due to a lack of systematic observations spanning gradients of drainage area and climate (Kochel, 1988).

Previous work has suggested that climate etches its signature in river channels (Hartshorn et al., 2002; Molnar et al., 2006; Stark et al., 2010), but it is not clear over what time scales this imprint develops, or how precipitation changes are reflected in alluvial riverbed elevation. Precipitation generates runoff and erodes sediment from various parts of a drainage basin; a proportion of it forms the riverbed, with grain sizes that provide resistance to subsequent transport. This bed sediment is transported through the river network by streamflow once it exceeds the entrainment threshold. The frequency of such transport is affected by climatic regimes (Groisman et al., 2001; Molnar et al., 2006; O’Connor and Costa, 2004; Pitlick, 1994), producing reach-scale scour/fill (short-term variability) or aggradation/degradation (long-term trends) over multiple storm cycles. For steady basin conditions, changes in riverbed elevation reflect the balance between supply and transport capacity, i.e., net sediment flux through the fluvial system.

Over longer time scales (104–106 yr), net sediment flux and fluvial erosion have been generalized using landscape-evolution models to assess climate impacts on river incision and drainage network development (e.g., Tucker and Slingerland, 1997). These models provide insight on the long-term impacts of increasing humidity/aridity on channels, but are not appropriate for understanding the short-term responses of alluvial riverbeds to changes in climate, which are transient over their time domains. Other research has correlated modern records of precipitation (P) or discharge (Q) with vertical (Hartshorn et al., 2002) and lateral (Stark et al., 2010) erosion rates in bedrock rivers, but the spatial and temporal mismatches between P, Q, and erosion data sets make it difficult to interpret transient responses of river systems to climate. Likewise, concurrent paleorecords of climate and sedimentation (103–105 yr) have been employed to provide inferences of past landscape response (Knox, 2000). However, such analyses are complicated by temporal biases in deposited sediments (Jerolmack and Paola, 2010; Sadler, 1981), poorly constrained prehistoric climate records, and feedbacks between riverbed elevation and sedimentation rates (Daniels, 2008).

In the United States there is excellent documentation of changes in the magnitude and variability of P (Karl et al., 1995), alteration of its timing and form (Knowles et al., 2006), and trends in the frequency of extreme rainfall and associated floods (Groisman et al., 2001) for the past several decades. Yet despite decades of P and Q measurement, the links between climate and contemporaneous changes in river morphology remain unexplored at broad spatial scales. This shortcoming thwarts short-term prediction of how climate change might affect flood risk (e.g., ratios of flood stage to floodplain elevation), longitudinal sediment storage, and the spatial distribution and stability of aquatic habitat (e.g., frequency of bed disturbance events and amount of in-channel sediment storage), as well as long-term understanding of drainage basin development (e.g., river longitudinal profile shape).

In this paper, we examine relationships between riverbed elevation time series, basin characteristics, climate, and climate change. Our purpose is to (1) investigate broad-scale controls on riverbed elevation trends; (2) assess the variability in this signature across climatic gradients; and (3) determine how these trends may be affected by climate changes.

We analyzed publicly available U.S. Geological Survey streamflow measurements and daily flow data for reference (i.e., a classification that denotes watersheds that are minimally disturbed by human influences) gaging stations in the Gages-II data set (Falcone, 2011) for 1950–2011. Mean channel bed elevation (BE) was calculated from each Q measurement as: BE = h − (A/w), where h is flow stage above local datum, A is flow area, and w is flow width, assuming a rectangular cross section. The values for h, A, and w are provided for each measurement in the database and we only utilized measurements from stable gage sites since 1950. Trends were investigated by evaluating these calculated BE values for individual stations over time. For gaging sites with time series trends (bed elevation trend, BET) in BE (p < 0.05), we derived annualized rates of aggradation/degradation. For those without significant trends, we calculated the interquartile range of BE (bed elevation variability, BEV). We also computed annual Q percentiles (Q90 and Q50 in m3/s) from daily streamflow data and temporal trends (p < 0.05) in these percentiles (Q90trend and Q50trend in m3/s/yr).

We assessed BET and BEV with respect to Q90, Q50, Q90trend, Q50trend, a measure of high-flow variability (Q90/Q50), mean annual precipitation (meanP in mm), annualized trend (p < 0.05) in mean precipitation (Ptrend in mm/yr) (Mitchell and Jones, 2005), as well as drainage basin characteristics that are generally assumed to control river incision, including drainage area (DA in km2), site elevation (m above sea level, masl), mean watershed slope (%), drainage density (km/km2), and dominant basin lithology (see the GSA Data Repository1).

We find decadal trends of bed elevation change for 68% of the 915 stations analyzed. Of these, riverbeds were aggrading (BET+) at 38% and degrading (BET−) at 62%. Figures 1A and 1B show a broad mix of BET and BEV sites, respectively, with no obvious systematic geographic biases. In addition, there are regions where BET+ and BET− sites are colocated, as well as local concentrations of aggrading (Pacific Northwest) and degrading (Appalachian Mountains) stations within similar meanP regimes, but these also occur in regions that have had contrasting trends in meanP (Fig. 1B; Groisman et al., 2001).

Figure 1.

A: Map of bed elevation trend (BET) against background of mean annual precipitation (meanP) values (see the Data Repository [see footnote 1]). B: Map of bed elevation variability (BEV) against background of annualized trend in mean precipitation (Ptrend) values. C: Time series at site 05458000: daily discharge (Q, m3/s), annual Q percentiles (Q50 and Q90, m3/s), and BE (m) (for statistics, see the Data Repository). Dashed lines represent interquartile range of BE, used in calculation of BEV, and red line marks degrading BE trend. D: Boxplots of aggradation (BET+), degradation (BET−), and BEV versus drainage area. Boxes, depicting distributional interquartile range, are variable width based on number of stations in each bin. Color stripes indicate proportionality of sites within each category. Values to right of each box indicate distribution median (black) and n value (blue). Mean number of measurements per site was 220 for BET+/– and 163 for BEV sites, with 34 yr and 30 yr of data, respectively. Overall, decadal median values are +0.5 cm/yr for BET+ and −0.6 cm/yr for BET–, considerably higher than most documented long-term, basin-averaged rates of erosion in mountain belts and short-term rates derived from suspended sediment yields (Kirchner et al., 2001) (ASL—above sea level).

Figure 1.

A: Map of bed elevation trend (BET) against background of mean annual precipitation (meanP) values (see the Data Repository [see footnote 1]). B: Map of bed elevation variability (BEV) against background of annualized trend in mean precipitation (Ptrend) values. C: Time series at site 05458000: daily discharge (Q, m3/s), annual Q percentiles (Q50 and Q90, m3/s), and BE (m) (for statistics, see the Data Repository). Dashed lines represent interquartile range of BE, used in calculation of BEV, and red line marks degrading BE trend. D: Boxplots of aggradation (BET+), degradation (BET−), and BEV versus drainage area. Boxes, depicting distributional interquartile range, are variable width based on number of stations in each bin. Color stripes indicate proportionality of sites within each category. Values to right of each box indicate distribution median (black) and n value (blue). Mean number of measurements per site was 220 for BET+/– and 163 for BEV sites, with 34 yr and 30 yr of data, respectively. Overall, decadal median values are +0.5 cm/yr for BET+ and −0.6 cm/yr for BET–, considerably higher than most documented long-term, basin-averaged rates of erosion in mountain belts and short-term rates derived from suspended sediment yields (Kirchner et al., 2001) (ASL—above sea level).

BE time series reveal that high-frequency fluctuations in riverbed level are typically caused by short-term variability in local hydrology (expressed as BEV at otherwise stable sites; Fig. 1B). Over decadal time scales, however, BE trends capture systemic shifts in climate and/or sediment supply (e.g., decreasing BE, concomitant with an upward shift in Q50 and Q90; Fig. 1C). These factors suggest that changes in bed elevation may be intrinsically linked to spatial and temporal differences in the frequency of sediment-transporting events expressed between climatic regions (Baker, 1977; Knowles et al., 2006; Molnar et al., 2006; Tucker and Bras, 2000).

Drainage Basin Controls

Models of long-term landscape evolution often assume constant basinwide sediment supply, so stream power (e.g., Q*channel gradient) controls rates of bed elevation change (BEC) and further, Q scales with DA. We find that none of the following variables correlates with the magnitude of BET or BEV: mean watershed slope, drainage density, relief, elevation, meanP, or lithology (Fig. DR1 in the Data Repository). However, our results show that the magnitude (not direction) of decadal trends in riverbed elevation (both BET+ and BET−) is correlated with DA, Q50, and Q90 (Fig. 1D; Fig. DR1), indicating a positive first-order relationship between the magnitude of BET and BEV and the volume of erodible alluvial fill (and thus scour depth) in progressively larger basins, regardless of network structure, basin physiography, runoff-generating mechanisms, or geologic materials.

Imprint of Climate

Building on prior work (Molnar et al., 2006; Pitlick, 1994; Tucker and Slingerland, 1997), we investigate the hypothesis that the flood events capable of inducing long-term change in riverbed elevation are largely a function of regionally determined P variability (Tucker and Bras, 2000). Climatically, this should be expressed in Q distributions as a ratio between high and mean flows. Thus, we computed Q90/Q50, or high flow variability, as a proxy for riverbed shear-stress distribution above a sediment transport threshold (Baker, 1977; Molnar et al., 2006; Pickup and Warner, 1976). We truncated this distribution at Q90 in order to minimize the influence of rare floods, and normalized BET+ and BET− values by the median depth from all Q measurements for each site, in order to investigate the climatic expression of Q on BET in a scale-independent way.

We find monotonically increasing relationships between depth-normalized BET+ and BET− and Q90/Q50 (Fig. 2A), consistent with the idea that high flow variability controls sediment flux and morphologic change (Molnar et al., 2006; Tucker and Slingerland, 1997). Bed material flux occurs at ∼Q50 (depending on climatic regime) as transport of the active sediment layer (Emmett and Wolman, 2001), and on the high end of the Q range as reshaping of the channel (Pickup and Warner, 1976). Therefore, we may assume that stations with higher Q90/Q50 are prone to more frequent channel scour/fill, subject to constraints of an erodible riverbed, leading to greater potential for BET. Stations with high Q90/Q50 and with positive Q90/Q50 trends might be expected to have greater potential for BE trends. [There are no systematic bed-material data available to test the importance of grain size on BET and BEV, but proxy data on basin lithology correlate poorly with BEC (Fig. DR1), suggesting that rock type is not a first-order control on BE in alluvial channels.]

Figure 2.

A: Bed elevation trend (BET) normalized by the median flow depth from all streamflow measurements at each site versus high-flow variability ratio (Q90/Q50; annual Q percentiles, where Q = discharge). B and C: Mean annual precipitation (meanP) versus Q90/Q50. Symbols as in Figure 1D.

Figure 2.

A: Bed elevation trend (BET) normalized by the median flow depth from all streamflow measurements at each site versus high-flow variability ratio (Q90/Q50; annual Q percentiles, where Q = discharge). B and C: Mean annual precipitation (meanP) versus Q90/Q50. Symbols as in Figure 1D.

Although the relationship between rates of BET+/− and meanP is poorly defined (Fig. DR1), we find that meanP has an indirect influence on BEC. Q90/Q50, which is directly correlated with BET+ and BET− (Fig. 2A), is inversely proportional to meanP (i.e., Q90/Q50 decreases systematically with increasing meanP) (Figs. 2B and 2C), highlighting the variable expression of climate in Q distributions. Dry regions have higher Q90 for the same Q50 than wet regions, consistent with prior work (Turcotte and Greene, 1993). However, even though Q90/Q50 is typically lower at wetter sites (meanP > 1000 mm), Q50 is apparently sufficiently high above the threshold for sediment flux at these sites to enable high values of change.

Impacts of Changes in Climate

Given the monotonic relationships between Q90/Q50 and BET/BEV, and the corresponding inverse one between Q90/Q50 and meanP (Fig. 2), we question the potential impact of observed changes in P and Q on BET and BEV. To remove any scale bias we computed percentage changes in Q, and we find significant Qtrends at 26% of sites for which sufficient data exist (n = 187) that reflect documented Ptrends for the contributing watersheds (Mitchell and Jones, 2005). Qtrends (Q50 and Q90) increase with Ptrend to 2 mm/yr, and negative Ptrends are associated with concomitant declining trends in both Q percentiles (Fig. 3A). We also find that moderately positive Ptrends (1–2 mm/yr) are disproportionately expressed as increasing Qtrends in relatively dry watersheds, or those with high values Q90/Q50, yet the highest Ptrends are not associated with the highest Qtrends (Fig. 3A), suggesting there is asymptotic buffering of extreme changes in rainfall in wet basins (e.g., via floodplain storage).

Figure 3.

Annual percentage change in temporal trend in annual Q (i.e., discharge) percentiles (Q50trend and Q90trend) normalized by median value of Q for each site. A: Versus annualized trend in mean precipitation (Ptrend). B: Versus mean annual precipitation (meanP). Within each x-axis category, Q50trend is represented by left box and Q90trend by right box. Symbols as in Figure 1D. We find a similar number of sites with increasing and decreasing Q90trends (n = 47:42), but twice as many sites with increasing Q50trends (n = 103:49).

Figure 3.

Annual percentage change in temporal trend in annual Q (i.e., discharge) percentiles (Q50trend and Q90trend) normalized by median value of Q for each site. A: Versus annualized trend in mean precipitation (Ptrend). B: Versus mean annual precipitation (meanP). Within each x-axis category, Q50trend is represented by left box and Q90trend by right box. Symbols as in Figure 1D. We find a similar number of sites with increasing and decreasing Q90trends (n = 47:42), but twice as many sites with increasing Q50trends (n = 103:49).

Furthermore, trends in both Q percentiles (Q50trend and Q90trend) are inversely related to meanP, such that Qtrend strength progressively declines in increasingly wet watersheds (Fig. 3B). This translates as an overall increase in Q in dry regions (median +1.1% and +1.3% per year for Q50trend and Q90trend, respectively, for sites with meanP < 600 mm), and a modest decrease at the wettest locations (median −0.5% and −0.6%, respectively, for those with meanP > 1400 mm), compared to the overall median value of either Qtrend across all climates (∼+0.7%). However, decreases in Q50 are also disproportionately expressed at dry sites and at high elevation (>1000 masl), and decreases in Q90 prevail disproportionately at wet sites (>1400 mm) (Fig. 3B). These disproportionate results suggest the complexity of climate change expression in streamflow.

How might such Qtrends affect bed elevation? Because the data sets are concurrent, it is challenging to discern cause from effect, especially if bed elevation response lags the climate change. However, we find that trends in both Q and BE occur at 18% of sites with Q data, and that marked increases and decreases (>+1% and <−1%) in Q50 and Q90 are both associated with higher rates of BEC (Table DR3 in the Data Repository). Humidification in both dry and wet climates is expressed through increasing trends in Q50 and/or Q90 (increases >+1%; Q50trends are more common). The upward shift in median Q, which is typically greater than increases in Q90, is consistent with decreasing Q90/Q50. This increases the flow duration above a fixed sediment flux threshold, potentially rendering the flow regime more geomorphically effective, regardless of small contractions in Q90/Q50. Aridification is typically expressed through declines in Q90, yet expansion of Q90/Q50 in dry and wet regions, and corresponds to high rates of BEC (see the Data Repository). At wet sites, one would expect aridification to reduce the already high number of sediment flux days, yet a small change is unlikely to have a major impact on BET and BEV, unless a transport/storage threshold is crossed. At arid sites it is plausible that changes in the flood frequency-magnitude relationship, especially in Q90, may affect BEC over much longer time scales, since bed elevation is essentially a function of larger and less frequent channel-shaping floods, but further work is required to clarify this effect.

We document climatically influenced decadal BE trends and variability across the continental United States. We demonstrate that Q variability varies by climatic regime proportional to BEC. Climate change, affecting the variability in high Q in contrasting climates, may act to redistribute the prevailing signature of climate in alluvial riverbeds. We find nonstationarity in BE at most sites, which has important implications for assessing flood risk, aquatic habitat, and stability of infrastructure, as well as for channel sediment budgets and longitudinal profiles, and the overall impact of climate change on fluvial systems.

This work was partially supported by a Natural Environment Research Council Ph.D. studentship to Slater. We thank T. Raub, P. Cawood, K. Juracek, J. Falcone, E. Gabet, K. Michaelides, J. Pizzuto, T. Dunne, J.W. Kirchner and several anonymous reviewers for helpful discussions and comments that improved the manuscript.

1.
ASCE Task Committee on Sediment Transport and Aquatic Habitats, Sedimentation Committee
,
1992
,
Sediment and aquatic habitat in river systems
:
Journal of Hydrologic Engineering
 , v.
118
, p.
669
687
,
doi:10.1061/(ASCE)0733-9429(1992)118:5(669)
.
2.
Baker
V.R.
,
1977
,
Stream-channel response to floods, with examples from central Texas
:
Geological Society of America Bulletin
 , v.
88
, p.
1057
1071
,
doi:10.1130/0016-7606(1977)88<1057
.
3.
Daniels
J.
,
2008
,
Distinguishing allogenic from autogenic causes of bed elevation change in late Quaternary alluvial stratigraphic records
:
Geomorphology
 , v.
101
, p.
159
171
,
doi:10.1016/j.geomorph.2008.05.022
.
4.
Emmett
W.W.
Wolman
M.G.
,
2001
,
Effective discharge and gravel-bed rivers
:
Earth Surface Processes and Landforms
 , v.
26
, p.
1369
1380
,
doi:10.1002/esp.303
.
5.
Falcone
J.A.
,
2011
,
GAGES-II: Geospatial attributes of gages for evaluating streamflow: Digital spatial dataset
:
Reston, Virginia, U.S. Geological Survey
 , p.
1
24
, .
6.
Gregory
K.
,
2006
,
The human role in changing river channels
:
Geomorphology
 , v.
79
, p.
172
191
,
doi:10.1016/j.geomorph.2006.06.018
.
7.
Groisman
P.Y.
Knight
R.W.
Karl
T.R.
,
2001
,
Heavy precipitation and high streamflow in the contiguous United States: Trends in the twentieth century
:
American Meteorological Society Bulletin
 , v.
82
, p.
219
246
,
doi:10.1175/1520-0477(2001)082<0219:HPAHSI>2.3.CO;2
.
8.
Hartshorn
K.
Hovius
N.
Dade
W.B.
Slingerland
R.L.
,
2002
,
Climate-driven bedrock incision in an active mountain belt
:
Science
 , v.
297
, p.
2036
,
doi:10.1126/science.1075078
.
9.
Inman
D.L.
Jenkins
S.A.
,
1999
,
Climate change and the episodicity of sediment flux of small California rivers
:
Journal of Geology
 , v.
107
, p.
251
270
,
doi:10.1086/314346
.
10.
Jerolmack
D.J.
Paola
C.
,
2010
,
Shredding of environmental signals by sediment transport
:
Geophysical Research Letters
 , v.
37
, p.
1
5
,
doi:10.1029/2010GL044638
.
11.
Karl
T.R.
Knight
R.W.
Plummer
N.
,
1995
,
Trends in high-frequency climate variability in the twentieth century
:
Nature
 , v.
377
, p.
217
220
,
doi:10.1038/377217a0
.
12.
Kirchner
J.
Finkel
R.C.
Riebe
C.S.
Granger
D.E.
Clayton
J.L.
King
J.G.
Megahan
W.F.
,
2001
,
Mountain erosion over 10 yr, 10 k.y., and 10 m.y. time scales
:
Geology
 , v.
29
, p.
591
594
,
doi:10.1130/0091-7613(2001)029<0591:MEOYKY>2.0.CO;2
.
13.
Knowles
N.
Dettinger
M.D.
Cayan
D.R.
,
2006
,
Trends in snowfall versus rainfall in the western United States
:
Journal of Climate
 , v.
19
, p.
4545
4559
,
doi:10.1175/JCLI3850.1
.
14.
Knox
J.C.
,
2000
,
Sensitivity of modern and Holocene floods to climate change
:
Quaternary Science Reviews
 , v.
19
, p.
439
457
,
doi:10.1016/S0277-3791(99)00074-8
.
15.
Kochel
R.C.C.
,
1988
,
Geomorphic impact of large floods: Review and new perspectives on magnitude and frequency
, in
Baker
V.R.
et al
., eds.,
Flood geomorphology
 :
New York
,
J. Wiley & Sons
, p.
169
187
.
16.
Mitchell
T.
Jones
P.
,
2005
,
An improved method of constructing a database of monthly climate observations and associated high-resolution grids
:
International Journal of Climatology
 , v.
25
, p.
693
712
,
doi:10.1002/joc.1181
.
17.
Molnar
P.
Anderson
R.S.
Kier
G.
Rose
J.
,
2006
,
Relationships among probability distributions of stream discharges in floods, climate, bed load transport, and river incision
:
Journal of Geophysical Research
 , v.
111
,
F02001
,
doi:10.1029/2005JF000310
.
18.
O’Connor
J.E.
Costa
J.E.
,
2004
,
Spatial distribution of the largest rainfall-runoff floods from basins between 2.6 and 26,000 km2 in the United States and Puerto Rico
:
Water Resources Research
 , v.
40
, p.
1
11
,
doi:10.1029/2003WR002247
.
19.
Pickup
G.
Warner
R.F.
,
1976
,
Effects of hydrologic regime on magnitude and frequency of dominant discharge
:
Journal of Hydrology
 , v.
29
, p.
51
75
,
doi:10.1016/0022-1694(76)90005-6
.
20.
Pitlick
J.
,
1994
,
Relation between peak flows, precipitation, and physiography for five mountainous regions in the western USA
:
Journal of Hydrology
 , v.
158
, p.
219
240
,
doi:10.1016/0022-1694(94)90055-8
.
21.
Sadler
P.
,
1981
,
Sediment accumulation rates and the completeness of stratigraphic sections
:
Journal of Geology
 , v.
89
, p.
569
584
,
doi:10.1086/628623
.
22.
Simon
A.
,
1989
,
A model of channel response in disturbed alluvial channels
:
Earth Surface Processes and Landforms
 , v.
14
, p.
11
26
,
doi:10.1002/esp.3290140103
.
23.
Stark
C.
Barbour
J.R.
Hayakawa
Y.S.
Hattanji
T.
Hovius
N.
Chen
H.
Lin
C.-W.
Horng
M.-J.
Xu
K.-Q.
Fukahata
Y.
,
2010
,
The climatic signature of incised river meanders
:
Science
 , v.
327
, p.
1497
1501
,
doi:10.1126/science.1184406
.
24.
Syvitski
J.P.M.
Vo
C.J.
Kettner
A.J.
Green
P.
,
2005
,
Impact of humans on the flux of terrestrial sediment
:
Science
 , v.
308
, p.
376
380
,
doi:10.1126/science.1109454
.
25.
Tucker
G.E.
Bras
R.L.
,
2000
,
A stochastic approach to modeling the role of rainfall variability in drainage basin evolution
:
Water Resources Research
 , v.
36
, p.
1953
1964
,
doi:10.1029/2000WR900065
.
26.
Tucker
G.E.
Slingerland
R.L.
,
1997
,
Drainage basin responses to climate change
:
Water Resources Research
 , v.
33
, p.
2031
2047
,
doi:10.1029/97WR00409
.
27.
Turcotte
D.L.
Greene
L.
,
1993
,
A scale-invariant approach to flood-frequency analysis
:
Stochastic Hydrology and Hydraulics
 , v.
7
, p.
33
40
,
doi:10.1007/BF01581565
.
28.
Williams
G.P.
Wolman
M.G.
,
1984
,
Downstream effects of dams on alluvial rivers
:
U.S. Geological Survey Professional Paper
 
1286
,
83
p.
1GSA Data Repository item 2013163, supplementary methods, Figure DR1, and Tables DR1–DR4, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.